Hands-on lab

Prompt Workshop: Make AI RequestsClear

Test, revise, and evaluate prompts through hands-on experiments.

Start the lab See the syllabus

1 htotal time

3modules

5lessons

4self-checks

Syllabus

3 modules. One clear path.

Built for People who want more useful, reliable responses from generative AI tools

001 / Module
Build a Useful First Prompt
Give the AI a clear task, useful context, and a defined output.
1 lessons · 12 min
- 1. Turn a vague request into a specific one
002 / Module
Add Context and Boundaries
Supply relevant information, set limits, and avoid unnecessary detail.
2 lessons · 24 min
- 1. Give the AI the context it needs
- 2. Make instructions testable
003 / Module
Iterate and Check the Output
Use response evidence to revise prompts and verify important claims.
2 lessons · 24 min
- 1. Revise with one change at a time
- 2. Check before you rely on an answer

The lab

Learn by doing.

Pick any lesson from the tree. Flip the cards, answer the checks and mark each lesson complete to fill your progress receipt.

lesson.mdLesson 1/5

Try itModule 112 min

Turn a vague request into a specific one

Compare a broad request with one that gives the AI a practical target.

Identify the task, audience, and output format in a prompt.
Draft a prompt that includes all three.

A prompt is an instruction or question sent to an AI system. The system responds using the information and direction you provide, so vague requests often produce generic answers.

Start with the job you need done. Add who the result is for and what shape it should take. For example, ask for a short checklist for a new team member rather than simply asking for advice about teamwork.

Run the prompt test

1
Choose a simple job
Pick a low-risk task, such as explaining a household process or organizing study notes.
2
Write a vague prompt
Ask for help with the task in one short sentence. Save the response.
3
Add three details
Rewrite the prompt to name the task, audience, and output format.
4
Compare results
Note which response is easier to use and which prompt detail made the biggest difference.

Knowledge checkWhich prompt gives the AI the clearest target?ATell me about gardening.BExplain gardening.CWrite a five-item checklist for a beginner preparing a small balcony garden.

ReflectWhat changed between your two responses when you named the task, audience, and format?

Saved in this browser only. Nothing is sent anywhere.

Try itModule 214 min

Give the AI the context it needs

Test how a focused background note can improve a response.

Separate relevant context from distracting detail.
Add a constraint that makes an answer more usable.

Context is information that helps the AI interpret your request. It might include a reader's current knowledge, the situation, or a source passage to use. A constraint sets a boundary, such as a word limit or a requirement to use only the supplied facts.

More context is not always better. Include details that affect the answer, and state important limits plainly.

Run a context experiment

1
Choose a task
Ask the AI to draft a short explanation of a topic you know well.
2
Add audience context
Say what the reader already knows and what they need to do with the explanation.
3
Add one boundary
Set a practical limit, such as a maximum length or a request to avoid unexplained jargon.
4
Inspect the answer
Mark which sentences became more relevant and whether the boundary was followed.

ReflectWhich context detail improved the answer, and what detail could you remove without changing it?

Saved in this browser only. Nothing is sent anywhere.

Try itModule 210 min

Make instructions testable

Replace fuzzy preferences with criteria you can inspect.

Rewrite a vague instruction as an observable criterion.

Words such as clear, engaging, and professional can mean different things to different readers. Translate them into visible features. For example, replace “make it clear” with “use short sentences, define technical terms, and give one example.”

Sort it

Sort each instruction by whether it gives an observable criterion.

Use three bullets and keep each under 12 words.
Make it sound better.
Define each technical term on first use.
Make the tone nice.

Knowledge checkWhich instruction is easiest to verify in the response?AMake the explanation more polished.BUse a heading, then give three steps in numbered order.CMake the answer feel complete.

Try itModule 313 min

Revise with one change at a time

Diagnose a weak response and make a targeted improvement.

Identify a specific gap in an AI response.
Revise a prompt to address that gap.

Run a revision loop

1
Save the first prompt
Keep the original wording and response so you can compare versions.
2
Name one problem
Be specific: the answer may be too broad, omit a needed example, or ignore a format instruction.
3
Change one instruction
Add or clarify the prompt detail that addresses that problem. Leave other wording unchanged.
4
Compare versions
Check whether the target improved and whether the change caused any new issue.

ReflectWhat did you change in your prompt, and what evidence in the new response shows whether it helped?

Saved in this browser only. Nothing is sent anywhere.

Try itModule 311 min

Check before you rely on an answer

Treat fluent text as a draft, not proof that its claims are correct.

Identify claims that need independent checking.
Add a verification request to a prompt.

AI responses can sound confident while containing errors, missing context, or invented details. Check names, dates, quotations, calculations, and claims that could affect a decision. For important information, compare against trustworthy sources rather than relying on the response alone.

You can ask the AI to separate supplied facts from assumptions or to flag points that need verification. That can help you review the answer, but it does not verify the claims for you.

Knowledge checkA response includes a precise statistic with no source. What is the best next step?AUse it because precise numbers are usually reliable.BAsk the AI to repeat the statistic more confidently.CCheck the statistic against a trustworthy source and note its date and context.

ReflectWhich claim in your test response would you verify first, and what kind of source would be suitable?

Saved in this browser only. Nothing is sent anywhere.

Progress

Your receipt.

0/5

Lessons completed. Everything is tracked in this browser only. Nothing is sent anywhere.

Clear AI PromptsProgress

1 · Build a Useful First Prompt0/1

2 · Add Context and Boundaries0/2

3 · Iterate and Check the Output0/2

Time completed0 min

Checks answered0/4

Reflections saved0/4

Total0%

A demo receipt for a fictional course. It is not a credential.

Your guide

Mira Chen

Prompt design educator at Northstar Learning Studio (fictional)

Mira develops practical exercises that help new AI users communicate tasks, context, and constraints clearly.

Fictional guide and course, generated for this demo.

FAQ

Do I need a special prompt formula?

No single formula works for every task. Start with the task, audience, useful context, and desired output, then revise based on the response.

Should I put every detail into one prompt?

Include details that could change the answer. Keep the main task easy to find, and add constraints that you can check.

Can I trust an answer if the prompt is precise?

A precise prompt can improve relevance, but it cannot guarantee accuracy. Check consequential or factual claims against trustworthy sources.

Enrol

Ready to start?

1 h of self-paced learning across 3 modules. Free, and your progress stays on your device.

enrol --course CADemo

Write prompts that specify a task, audience, and desired format.Add relevant context and constraints without burying the main request.Revise a prompt using evidence from the AI response.Check generated answers for unsupported claims and missing details.Write prompts that specify a task, audience, and desired format.Add relevant context and constraints without burying the main request.Revise a prompt using evidence from the AI response.Check generated answers for unsupported claims and missing details.